The right intelligence for every great idea.

Deep reasoning model with careful multi-step strategy work and nuanced brand-safe writing.

Cinematic video with native audio, physics-accurate motion and strong prompt control.

OpenAI's flagship general model with strong writing, planning and instruction-following across long documents.

Photorealistic image model with exceptional detail, lighting and skin rendering.

Expressive voice synthesis with emotional direction, cloning and 70+ languages.

OpenAI's flagship video model with synced audio, strong physics and multi-shot narrative consistency.

Luma AI's unified multimodal model. Luma AI provides advanced image, video and multimodal capabilities that can be accessed through the DreamCache Model Hub, subject to API availability.

Google's highest-fidelity image model with excellent text rendering and prompt adherence.

Agentic coding model for building sites, prototypes and production tools from briefs.

High-fidelity instrumental music with fine-grained control over mood, tempo and instrumentation.

Long-context multimodal model that reads video, audio, images and documents in a single request.

Instruction-based image editing that preserves characters, products and scenes across edits.

Luma's video model sharing the Uni-1 foundation that powers DreamCache — natural motion, HDR and reasoning-guided direction.

Full-song generation with vocals, structure control and radio-ready mixing.

Vector-native design model producing brand-consistent SVG, icons and layout systems.

In-context video editing — restyle, relight, remove and extend existing footage with prompts.

Best-in-class text rendering inside images — posters, packaging, lockups and layouts.

Photoreal talking avatars with natural gesture and 175+ languages from a single photo.

Instruction-following image generation and editing with best-in-class text rendering and layout awareness.

Gemini-powered image editing with uncanny character consistency, scene blending and conversational refinement.

State-of-the-art speech-to-text with word-level timestamps, diarization and 99-language coverage.

Search-grounded answers with citations, tuned for factual research and source-backed briefs.

Open reasoning model with near-frontier chain-of-thought quality at a fraction of the cost.

Fast, accurate transcription with diarisation and keyword boosting at very low cost.

Text- and image-to-3D with clean topology, PBR textures and game-ready export.

Agentic model tuned for long tool-use chains, browsing and multi-step task execution.

Multilingual model with top-tier quality across 100+ languages for localisation at scale.

Very fast image generation with strong aesthetics, tuned for social-scale content volume.

Real-time knowledge model with live web and X data for current-events research and trend tracking.

Fast, affordable video generation with reliable motion for social-first formats.

Ultra-low-latency streaming TTS built for real-time voice agents and live experiences.

Luma's Dream Machine platform API powered by Ray 3 — fast, cinematic text- and image-to-video with strong camera and physics control.

Commercially-safe image generation trained on licensed content, deeply integrated with Creative Cloud.

Europe's frontier language model with strong multilingual output, EU data residency and open deployment options.

Retrieval-optimised model with first-class embeddings and grounded RAG answers over brand knowledge.

Budget-friendly multi-shot video with solid coherence for high-volume social pipelines.

Fast sound-effect and ambience generation with precise duration control at low cost.

Open-weight workhorse with strong quality-per-dollar, deployable on any inference provider.

Llama 4 served on Groq LPUs at extreme token throughput for real-time drafting and chat.